Translations: As of 9-15-04, the President's campaign staff and conservative 527's had done a significantly better job of registering voters than Kerry's staff and liberal 527's.
Note that this analysis takes advantage of two assumptions which might be problematic.
First, we are assuming that people who voted in 2000 will not change their vote in 2004. In other words, we are assuming a stable electorate. I do not think the electorate will be stable, but I do not think it will swing toward Kerry. Can Kerry do better than Gore did in 2000? I do not think so. He might do better than Gore by registering more people to vote for him...but that is the dependent variable I am testing here. Thus, if this assumption is inaccurate, it will be to Bush's benefit.
Second, I am assuming that the people who have left the voter registration rolls break down for Bush or Kerry the same way that all registered voters broke down for Bush or Gore in 2000. Thus, Bush enjoyed 7.72% support among registered voters in Wabasha County. I am assuming that of the registered voters who have been dropped from the rolls, Bush would have enjoyed 7.72%, Kerry would have enjoyed 32.51%, and roughly 3/5ths of those voters would have supported somebody else or would have supported nobody. Now, I recognize that, for any given county, this figure is somewhat biased. The chances are that people who have been dropped from the rolls in Wabasha County are more likely to be people in that 3/5ths group.
However, as long as we can assume that the people dropped from the rolls are not more likely to be Bush voters than Kerry voters, we can still conclude that Bush has gained ground on Kerry. Thus, for instance, in Wabasha County, as long as the number of Kerry people dropped were 4 times more numerous than the Bush people, our conclusion still holds that the president has gained ground in Wabasha County. Kerry received four times as many votes as Bush did in Wabasha. If Kerry lost 4 times as many voters in the cleaning of the registration rolls there, the overall conclusion about the county will hold (although we cannot identify a figure precisely; nevertheless, we can use the real figured mentioned above as rough estimates). So it goes for the entire state. I cannot think of any reason why, for any of these counties, we would expect anything beyond this. If Bush won a county with 5 times the number of votes Gore received, I think we can safely assume that he has lost 5 times as many voters to declining registration rolls.
Thanks for posting. I like this kind of analysis based on hard numbers and extrapolation. I think I will be checking this blog more often for this kind of interesting posting.